Fig. 5
From: On reliability of annotations in contextual emotion imagery
Examples of EMOTIC dataset images. For the left image, the following was obtained: 1 vote for Anger/Rage, 1 for Engagement/Anticipation, 1 for Disgust/Disconnection, 23 for Joy/Affection, 3 for Surprise/Amazement, 1 for Trust/Peace, and 0 for all other categories. For the right image, the next scores were obtained: 1 vote for Anger/Rage, 13 for Engagement/Anticipation, 7 for Disgust/Disconnection, 5 for Surprise/Amazement, 2 for Trust/Peace, and 0 for the remaining classes. It is to note the dispersion in the number of votes, which could be caused by the projection of emotions by annotators.
